JVM (2) garbage collection

Garbage collection

1、 How to judge whether the object is garbage

• Reference counting （ There is a reference to this object , Counting plus one 《 There are circular references 》）
• Reachability analysis algorithm （ To determine whether or not Root object GC Root Direct or indirect reference to 《 Grape kebabs 》）
• It can be used as GC Root The object of
  • Virtual machine stack （ Local variables in stack frames ） Object referenced in .
  • Object referenced by a class static property in a method area
  • The object referenced by a constant in the method area
  • Local method stack JNI（ In other words Native Method ） Referenced object

Four quotations ：

• Strong citation
• Soft citation （ After garbage collection, soft reference objects will be recycled again when memory is still insufficient ）
• Weak reference （ Garbage collection , Will recycle weak reference objects ）
• Virtual reference （ Mainly with ByteBuffer Use , Free direct memory ）

2、 Garbage collection algorithm

1） Mark clear

Definition ：Mark Sweep

• Faster
• Memory fragmentation will occur

2） Tag to sort out

Mark Compact

• Slow speed
• No memory fragmentation

3） Copy （ The new generation —>98% It's all life and death ）

Copy

• There will be no memory fragmentation
• It takes up twice the memory space

4） Generational recycling

• The newly created object is first assigned to eden District
• When the new generation is short of space , Trigger minor gc ,eden District and from Use of live objects - copy Copied to the to in , Living objects Life plus one , Then exchange from to
• minor gc May trigger stop the world, Pause other threads , After garbage collection , Resume user thread operation
• When the object in the surviving area Life exceeds the threshold 15 , Will be promoted to the old age
• When the old days were short of space , Will trigger minor gc, If space is still insufficient , Then trigger full gc ,stop the world Longer time ！

3、 Garbage collector

1) Serial The collector ( Serial )

characteristic ： Single thread 、 A simple and efficient （ Compared with the single thread of other collectors ）, Using replication algorithm . For a limited single CPU In terms of environment ,Serial The collector has no overhead of thread interaction , Focusing on garbage collection can naturally achieve the highest single thread collection efficiency . When the collector is recycling , All other worker threads must be suspended , Until it's over （Stop The World）！

2) Parallel GC The collector ( High throughput )

characteristic ： Belong to the new generation collector is also the collector using replication algorithm （ Used the surviving areas of the Cenozoic ）, again parallel The multithreaded collector of （ And ParNew Collector similar ）

The goal of this collector is to achieve a controllable throughput . There is also a point worthy of attention ：GC Adaptive adjustment strategy GCTimeRatio（ And ParNew The most important difference between collectors ）

3）CMS The collector （ Low latency ）

Concurrent Mark Sweep, A method aiming at obtaining the shortest recovery pause time Old age collector
characteristic ： be based on Mark - eliminate Algorithm implementation . Concurrent collection 、 Low pause , But will Generate memory fragmentation

CMS The collector runs as follows 4 Step ：

• Initial marker Mark GC Roots Objects that can be reached directly
• Concurrent Tags Conduct GC Roots Tracing The process of
• Re label （ Long time ） In order to fix the tag record of the part of the object whose tag changes because the user program continues to run during the concurrent tag ==（ Scan the entire heap memory ）==
• Concurrent elimination Clean up and recycle marked objects , In the process of clearing , There may still be new waste , This garbage is called floating garbage , If the user needs to store a large object , The new generation can't let go , In the old days, there was too much floating garbage , Will degenerate into serial Old The collector

CMS Collectors have three disadvantages ：

• Yes CPU Resource sensitive （ Complicated with common diseases ）
• Can't handle floating garbage —> Degenerate to serial Old The collector
• Mark Sweep —> Generate memory fragmentation —> Full gc

CMS The collector's memory recovery process is executed concurrently with the user thread , Collocation ParNew The collector （ Multithreading , The new generation , Copy algorithm ） And Serial Old The collector （ Single thread , Old age , Mark - Sorting algorithm ） Use .

4）G1 （First） The collector （ Low latency ）

characteristic ： A change in thought .G1 Will heap memory “ break up the whole into parts ”, Divide the heap memory into separate areas of equal size Region.（ Make it KPI（ Require at any time 1 In seconds ,STW The pause shall not exceed 10ms）.G1 Will try to achieve this goal , It can calculate the general area of this collection in reverse , Complete the collection incrementally .）

• No memory fragmentation （ Mark - Arrangement + Partial view replication ）
• Predictable pauses
• establish Region List of recycling values , Every time == Priority recovery is the most valuable （First）==Region
• Every Region Corresponding to one Remembered Set：
  • Yes Region Reference type data write operation in , produce Write Barrier barrier , Check whether the reference object is in another Region in ：
    • if , Through the card watch CartTable Record reference information to the referenced object Region Of Remembered Set in
  • To avoid When object references are different Region Full heap scan of objects

G1 The operation process of is similar to CMS broadly , There are four steps ：

• Initial marker Mark only GC Roots Objects that can be directly related to
• Concurrent Tags GC Root Start the reachability analysis of objects in the heap
• Final marker A brief treatment of the remains of Missing label object
• Screening and recovery ： each Region Sorted by recovery value and cost , Make a recycling plan based on the user's expected pause time , Sure Choose as many as you like Region Make up a recycling set , And then the part that decided to recycle Region Of live objects copied to empty Region in , And clean up the whole old Region The whole space of . The operation here involves moving the living object , It is necessary to pause the user thread , Done in parallel by multiple collector threads .

4、GC tuning

Tuning ideas ：

• Choose the right garbage collector
• Consider whether unnecessary objects are loaded into memory
• Is there a memory leak （ soft 、 Weak reference , cache ）
• The priority is to start in the Cenozoic GC（ Increase the proportion of Cenozoic generation 25%-50%）
• The new generation time is mainly spent in replication

Tuning case ：

• GC frequent ： enlarge New generation memory , Raise the promotion threshold ;
• （CMS） A single GC A long pause ： Recycle once in Cenozoic before re marking ;
• （jdk7） Old age is abundant FullGC： Increase permanent generation .